Industrial and domestic waste from the Cikakembang River heavily contributed to the pollution of Citarum River, yet pollution control is hindered by scarce monitoring data. Time-consuming trial-and-error methods often relied to estimate effluent from numerous outfalls. This study applied Bayesian reverse modeling to estimate wastewater discharges from 16 outfalls points under data-scarce conditions. The present study compared the model performance when constrained to hydraulic data versus water quality data. The hydraulic-based approach integrated DREAM with HEC-RAS unsteady flow simulation constrained to the synthetic water stages downstream, while the water quality approach combined DREAM with advection–dispersion equations model within MATLAB constrained to the observed Dissolved Oxygen (DO). Both approaches showed good performance, but differed in efficiency and uncertainty. Several multimodal posteriors were found in both approaches, representing the uncertainty rising from the scarce observation data. Computation wise, the hydraulic-constrained approach required tenfold greater computational time due to heavy load of HEC-RAS simulation, highlighting a trade-off between data availability of obtaining hydraulic data compared to water quality measurements. The findings provide policymakers with a practical tool for efficient resource allocation in data scarce river systems.
Yudianto et al. (Wed,) studied this question.